• Big Oilers Data II

    by Tyler Dellow • April 30, 2013 • Hockey • 11 Comments

    This is part of a series looking for reasons for the Oilers Corsi% collapse in 2012-13 by examining things on a shift-by-shift basis. Part 1 can be found here.

    In the first part of this series, I examined the performance of Oilers forwards in 2012-13 through three distinct lenses: the percentage of shifts on which they allowed a shot attempt against (SAA), the percentage of shifts with an SAA that turned into multi-SAA shifts and the percentage of shifts that were multi-SAA shifts. I did this in order to try and gain an understanding of how, at a shift-by-shift level, the performance of the Oilers forwards changed in 2012-13. Today, I’m going to examine the defencemen through this lens.

    First up is looking at the defencemen on the basis of the number of shifts with an SAA recorded. I’ll start with the guys who appeared both this year and last year.

    This graph is sorted, from left to right, by the size of the increase in the number of shifts with an SAA. If you’re Ryan Whitney’s agent, looking for a selling point, you can quite fairly say “When the Oilers defencemen went to hell last year in terms of shifts with an SAA, Whitney experienced the smallest growth in that number amongst anyone who took a regular shift!” Unfortunately, he was still really bad at this, just like he was in 2011-12. It’s just that other guys got worse.

    The Petry/Smid change really catches the eye. Those fellows were a pairing for most of the year – about 78% of Petry’s time on the ice at 5v5 was with Smid and about 81% of Smid’s 5v5 time was with Petry. There seemed to be a consensus that these guys weren’t as good this year as they were last year and this data would seem to provide some explanation for why people might have come to that conclusion. I’m not convinced that that’s true, but we’ll come to it.

    Potter experienced a rise in his percentage of shifts with an SAA this year. RELATED: Potter played more frequently with Whitney this year than he did last year. He had horrible numbers with him this year, just like he did last year.

    Nick Schultz, who played about 67% of his Oiler time in 2011-12 with Ryan Whitney, also saw a slight increase in his volume of shifts with an SAA this year, despite hardly playing Whitney. Why might that be? Well, let’s look at the numbers for the guys who weren’t on the Oilers for both seasons.

    Oh dear. Justin Schultz. It’s not perfectly clear but he saw at least a single SAA on 49.2% of his shifts in 2012-13. That’s an awful number. I hate to complain about things that are past but there might be a valuable lesson here about whether or not planning for a rookie defenceman to be in your top four when you’re a marginal playoff contender is a good idea. Fistric’s number is 44.3%, which isn’t great but, as with Potter, he shows significant Whitney effects.

    With no further comment, I note that Tom Gilbert doesn’t appear to have been such a bad performer in retrospect.

    Let’s switch lenses here and look at how the defencemen performed in terms shifts with at least one SAA that turned into multiple-SAA shifts. Again, the theory here is pretty straightforward: if you’re going to be on the ice for an SAA, you’d rather nip it in the bud, recover the puck and get out of the zone instead of continuing to be trapped in your zone while the opposition tees off. Again, I’ve sorted the graph from left to right, from biggest decrease to biggest increase.

    Welp. Not much for Ryan Whitney’s agent in this. In addition to continuing to be awful in terms of the percentage of shifts on which he experienced an SAA, he saw a rather large increase in his percentage of multi-SAA shifts. He was more likely than any non-Justin Schultz defenceman to be on the ice for a shift on which there was an SAA and he was the most likely of any of the defencemen to be on the ice for a multi-SAA shift.

    Petry and Smid jump out to me again. Way back in the day, Vic Ferrari made a fine point about defencemen who subsequently turned up in the All-Star game being more likely to have been given away at some point in their career. The canonical example of this is Larry Murphy, booed off a crappy team in Toronto, who went on to play just fine for a Stanley Cup champion in Detroit. It’s difficult to reconcile.

    One theory as to why this might be is that defencemen don’t have a ton of control over certain aspects of their results. If you’re on a team that spends a lot of time in its own end because it can’t generate opportunities at the other end of the ice, you’re going to look bad. Spending a lot of time in your own end of the ice is like hanging out in a bad part of town – the more time you spend there, the more likely it is that something bad will happen.

    You’ll recall that I pointed out above that Petry and Smid were more likely to be on the ice for a shift on which there was an SAA this year. I kind of theorized that the increased frequency with which this happened might explain why people thought that they looked poor – they were hanging out in a bad part of town. I find it awfully fascinating that the frequency with which those shifts turned into multi-SAA shifts didn’t really change from 2011-12 to 2012-13.

    In Petry’s case, it went up very slightly – from 40.2% to 40.3%. Effectively, he performed exactly the same as he did last year – 209 of his 518 shifts with at least one SAA were multi-SAA shifts. If he’d performed exactly the same as last year, it would be 208.5. It’s eerily similar. In Smid’s case, 196 of his 521 shifts with at least one SAA were multi-SAA shifts. If he’d posted the same percentage as last year (36.8% in 2011-12; 37.6% in 2012-13), he’d have had 191.5 multi-SAA shifts. To put that into perspective, it’s one extra shift that turns from a single SAA shift to a multi-SAA shift per 10.7 games.

    To underline this: Smid/Petry produced virtually identical performances in 2011-12 and 2012-13 once you limit the analysis to those shifts on which there was an SAA. To the extent that their defensive performance declined, it was in the volume of shifts on which there was an SAA, not the volume of multi-SAA shifts. This gives rise to a question, I think: to what extent do we attribute blame for that failing to Smid/Petry and to what extent is it due to things that were happening elsewhere?

    Corey Potter and Nick Schultz experienced moves in opposite directions with their percentages of multi-SAA shifts. In both cases, I think we can find some explanation of this in the data for guys who were new to the team this year.

    Mark Fistric and Justin Schultz were at opposite ends of the spectrum this year. Justin had shifts on which he allowed at least one SAA turn into multi-SAA shifts 42.1% of the time, the second worst percentage on the team. Fistric had a 35.3% number, which was the best on the team. Interestingly with Fistric, the Fistric/Potter pairing actually posted really good numbers this year – they were north of 50% as a Corsi%. The sample is small, they were a third pairing but that certainly wasn’t part of the problem for the OIlers.

    As far as Nick Schultz’s increase in his multi-SAA%, playing a lot with Justin probably didn’t help him. Justin was awfully weak at this; notably weaker than Whitney was last year. Again – it sure would have been nice if the Oilers had kind of planned for the rookie coming out of college to start in a third pairing role at 5v5 but c’est la vie.

    Let’s flip lenses again and look at multi-SAA shifts as a percentage of all shifts. As I mentioned in the last post, this sort of blends the ideas in the first two lenses. I tend to think it’s a bit less useful but, as they say, it is what is.

    I think I’ve set out above what, specifically, accounts for the increases in these numbers this year. Petry and Smid saw a greater share of their shifts lead to at least one SAA, while maintaining their multi-SAA number when you look only at shifts on which there was at least one SAA. Nick Schultz met Justin and things did not go that well defensively. (Probably noteworthy: Nick Schultz posted a better Corsi% when he wasn’t playing with Justin than when he was. If I had to guess, I’d guess that it had to do with Justin’s apparent weakness at preventing SAA and multi-SAA shifts.) Potter basically came out of things as a wash, a mix of playing with the guy with the worst numbers (Whitney) and the best (Fistric). For all the praise that Justin Schultz is getting – I think he’s going to finish ahead of Yakupov in Calder voting – he doesn’t show particularly well in these numbers.

    I want to talk a little about the idea of finding individual metrics that matter. Ultimately, what the people who are interested in analytics are doing, or trying to do, is find metrics that tell us about the individual who is on the ice. The objection of the “Hockey isn’t baseball” people is overstated in my opinion, but there’s merit to it. Every time a hockey player is on the ice, there are eleven other guys busy polluting his numbers. Right now, we try to infer whether or not a guy is contributing by looking at his top line underlying numbers, like Corsi% and then looking for things that might make us think that that’s not an adequate reflection of his numbers – maybe he’s getting really good or really poor ZoneStarts, or maybe he’s got strong/weak teammates. It’s alright as far as it goes – I tend to think it’s a decent way of looking at things – but it’s an awfully subjective sort of thing.

    What we really need to do is figure out how to isolate the things that a player does or doesn’t do that drive the Corsi%. That’s the trick. Thinking about defencemen this way strikes me as having some possibilities there. I’m not entirely convinced that defencemen are able to exercise a great deal of control over the number of shifts that they play on which the other team gets possession and is able to attack. If you play with Ryan O’Reilly, the puck’s going to be in the other end of the ice more than if you play with an NHL team foolish enough to give me minutes at centre ice.

    In that circumstance, you’d be the same guy, but the volume of shifts on which you sustained at least one SAA is going to be much higher, because I can’t keep the puck in the other end of the ice like Ryan O’Reilly can. If we think about this in terms of identifying a metric, what we really need is something that measures a defenceman’s performance during the opportunities that he does have to prevent a shift with an SAA from occurring. Keeping the play in the offensive zone once it’s there, to the limited extent that defencemen can do that. Breaking up plays that are coming towards your end of the ice before they result in an SAA. Turning the puck back around. When that sort of information and data starts to become available, I think we’re going to start seeing really good individual metrics for defencemen get developed.

    That’s kind of why I find looking at multi-SAA shifts/SAA shifts so interesting. We’re kind of starting everyone from a similar starting point there – we know that there’s been an SAA and then seeing how well they stop the bleeding. I would expect that the influence of the wings is less there than it is with the percentage of shifts with an SAA, because you’re kind of cutting out the impact of a forward line that just keeps things at the other end of the ice. You are, quite possibly, distilling more of the influence of the two defencemen and the centre, in terms of recovering the puck and getting out of the defensive zone. I don’t deny that the wingers are involved in that, I only suggest that the weighting of the involvement as between the five players on the ice isn’t 20/20/20/20/20. At the very least, I’d suggest that the relative weighting of the defencemen’s influence on that is greater than on the issue of whether or not a puck stays in the offensive zone.

    I wasn’t surprised to see that the Oilers’ pairing which was together last year and this year (Petry/Smid) essentially posted the same numbers at this. Armed with the knowledge of who the two worst Oiler defencemen were at preventing SAA and multi-SAA shifts, it’s probably unsurprising to find that Whitney and Justin Schultz were such an atrocious pairing. They played 120:06 together at 5v5 – two entire hockey games! – and posted an incredible 36.3% Corsi%. 36.3%! It’s just…wow.

    Up next, I’ll turn my attention towards the other end of the ice and SAF.

    Email Tyler Dellow at mc79hockey@gmail.com

    About Tyler Dellow

    11 Responses to Big Oilers Data II

    1. TrentonL
      April 30, 2013 at

      Can you break this down into when dubnyk or nk is in net? Looking at wowy the team is much worse in front of NK

      • Tyler Dellow
        April 30, 2013 at

        I doubt that there’s anything to it but I’ll give it a spin a bit later one, once I’m through this stuff.

    2. godot10
      April 30, 2013 at

      //I want to talk a little about the idea of finding individual metrics that matter. Ultimately, what the people who are interested in analytics are doing, or trying to do, is find metrics that tell us about the individual who is on the ice. The objection of the “Hockey isn’t baseball” people is overstated in my opinion, but there’s merit to it. Every time a hockey player is on the ice, there are eleven other guys busy polluting his numbers. Right now, we try to infer whether or not a guy is contributing by looking at his top line underlying numbers, like Corsi% and then looking for things that might make us think that that’s not an adequate reflection of his numbers – maybe he’s getting really good or really poor ZoneStarts, or maybe he’s got strong/weak teammates. It’s alright as far as it goes – I tend to think it’s a decent way of looking at things – but it’s an awfully subjective sort of thing.//

      Hypothesis: In hockey, one should invert the process. Instead of looking at an individuals’ stats to evaluate their worth, look at everyone else’s stats with and without the player, weighted by time on the ice together.

      i.e. To evaluate Whitney, look at everyone else with and without Whitney, and use that to assess Whitney’s contribution.

      For example, what is the average weighted difference in Corsi of Whitney’s teammates with Whitney on the ice vs. Whitney off the ice. Don’t look at Whitney’s Corsi to evaluate Whitney. Look at everyone else’s averaged weighted by ice time.

      Don’t look for direct measures. Use indirect measures.

      • Tyler Dellow
        April 30, 2013 at

        I think that’s what people are kind of doing when they got through the process I described, if not in the context of a formal formula. I’m more interested in finding direct measures.

    3. May 1, 2013 at

      Another excellent post Tyler. One thing I’d be interested in with this data is breaking things out into two categories, namely, shifts that begin with a DZ faceoff and shifts that don’t. It strikes me that shifts beginning with a DZ faceoff are much more likely to result in SAA and that they’ll sometimes include a sequence like “Lost FO – SAA – Puck Frozen – Lost FO – SAA”, and that a shift like that may well be a pretty successful shift for the defensemen.

      Another thing that might be worth controlling for is shifts that extend through a TV timeout (although maybe not since the effect is probably pretty small). It makes sense to me to count those as two different shifts (especially for your purposes here), but IIRC the NHL counts them as one.

      • Tyler Dellow
        May 1, 2013 at

        Scott -

        I’ll add it to the list. I’m thinking that I can poke into a lot of different areas with this data.

    4. Showerhead
      May 1, 2013 at

      Just some loose thinking here, but is it perhaps interesting that some of the guys who are particularly bad at giving up that 2nd shot are also the ones who, by eye, are bad at winning puck battles?

      Whitney was skating around with an anvil chained to his ankle, J Schultz fell off a predictable rookie-esque cliff, and N Schultz was tied to J Schultz. Interestingly enough, Smid and Petry are the sorts of guys you have a bit more faith in when it comes to going into the corner and pushing the puck in the right direction.

      It seems to me that the context you’re put in has a huge effect on that 1st shot against but your ability to win the ensuing puck battle is what controls the 2nd shot. I half wonder if your metric of shifts involving multiple shots against is one of the most noise-free defensive indicators I’ve stumbled upon.

      • Tyler Dellow
        May 1, 2013 at

        I…I kind of think that’s what I’m suggesting. ;)

        • Showerhead
          May 1, 2013 at

          LOL. So what you’re saying is I totally got it. I’m so smrt.

          • Ronald
            May 31, 2013 at

            , he ‘requested’ a trade, not demanded. Although in my mind, it’s pttery hard to get the words “Five year contract” out of my head–it is what he signed and agreed to; and he should at least have the decency to carry it out. I don’t want to diss Prongs and continue the ‘wear the pants not the panties’ mockery; but it makes me wonder. WHAT is going on?! Edmonton embraced Prongs and immediately we all loved him. He is a wickedly awesome defense player, one of the best in my opinion, and has never made any comments about not liking or being comfortable in Edmonton before. Which leads us to the only other factor which would cause him to suddenly leave Edmonton: the wife. Obviously she is of a rather spoiled background, and if all the ‘talk’ going around about her is true, then I must say I am extremely miffed at her. When you marry a man who plays for the NHL, the N stands for National. Not local. You have to accept that his career requires him to travel–if she wasn’t willing to stick it out and make it work, then why did she make the commitment? It is extremely selfish to make your husband do that to his career–it doesn’t just affect Chris, but the Oilers and all the Edmonton fans. If she misses her family and old town, then she’ll have to get used to it because he’s not going back to St. Louis. He makes a decent salary—take an airplane. You’ll have to do that from any other city he’s transferred to anyways!Edmonton is a wonderful city; I grew up there, and can think of no obvious differences there. Why the sudden hatred towards it? Is there really an affair going on between Pronger and a news anchor? I can’t say that Prongs struck me as that sort of person–he was always the solid, dependable player and carried that in person. Hardly the type to get another woman pregnant outside his marriage. But maybe I’m just in denial. When I saw how he contributed to our team, I got just as swept up in the Pronger-Mania as everybody else. So answer my questions, folks. Am I desperately clinging onto a would-be fantasy of a beloved defenseman with good reason? Or am a deluding myself in the fact that yes, he had a fling, and now he’s paying for it…? Does he deserve the benefit of doubt?I loved it while it lasted. And even though I’m only 16, I know a good player when I see one. At least Comrie and Weight will be able to show their faces in Edmonton again.

      • Jim
        May 1, 2013 at

        The 2nd shot/puck battles discussion rings true in light of a half-remembered posting of yours, Tyler, from some time ago. You advanced an argument that faceoffs were far less important than conventional wisdom holds, given (I think) the speed and frequency with which play travels from end to end and goalies high S% in most circumstances, and no doubt some other factors as well. Over the long term It is much more dangerous to be poor at gaining control of loose pucks and clearing them than to be losing defensive faceoffs, given most centres hover somewhere between 40 and 60% of faceoff wins. So why hand out contracts and tie up roster spots with face-off gunslingers who are ham-handed in the rest of the of hockey game?

        Unless you are the Oilers, then you are on the bad end of both…

        Not a great insight on my part with this remark, simply that your argument here helps put some other points of debate into place. This is a really useful set of posts. Best thing I’ve read this week (I’m marking 1200 pages of course work!). And speaking of which…

        This is a really provocative set of posts, th

    Leave a Reply

    Your email address will not be published. Required fields are marked *